{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp04_retrieval_first_paired20_20260802","cell_id":"invariant_mode_decomposition_design__medicine_healthcare","arm":"RETRIEVAL_FIRST","round_index":0,"hypotheses":[{"hypothesis_id":"H1","title":"Modal early warning for inpatient deterioration","problem":"Single-variable thresholds may miss coupled physiologic trajectories preceding acute deterioration.","affected_stakeholder":"Hospitalized patients, bedside nurses, and rapid-response teams.","workflow_boundary":"Four-hourly observation review through escalation or continued ward care.","failure_mode":"A weakly damped combination of vital signs, laboratory values, and oxygen support grows while each coordinate remains below its alert threshold.","unit_of_analysis":"Patient–four-hour state transition.","causal_lever":"Trigger a mode-matched assessment or care bundle when an adverse trajectory becomes unstable, not when an individual measurement crosses a limit.","archetype_mapping":"Estimate a local patient-state transition operator, decompose its modes, classify stability, and suspend modal alerts when residual error or mode drift exceeds tolerance.","expected_value":"Potentially earlier escalation at a fixed alert burden, contingent on adequate local stationarity and actionable modes.","falsifiable_claim":"At equal alerts per patient-day, the modal trigger will detect ICU transfer or emergency intervention at least two hours earlier than the incumbent score without lowering positive predictive value.","diversity_rationale":"Targets temporal physiologic stability at a patient-state boundary, unlike allocation, treatment-planning, dosing, or rehabilitation hypotheses.","mechanism_slugs":["eigendecomposition_workflow","modal_stability_analysis","spectral_gap_monitor"],"search_questions":["Have dynamic eigenmode or stability-based ward deterioration alerts already been evaluated?","Are successive inpatient measurements sufficiently dense and stationary for a useful local operator?","Which detected modes correspond to interventions clinicians can deliver promptly?"]},{"hypothesis_id":"H2","title":"Mode-targeted dialysis adjustment","problem":"Adjusting ultrafiltration from blood pressure alone may not prevent intradialytic hypotension caused by coupled volume, refill, cardiac, and autonomic responses.","affected_stakeholder":"Hemodialysis patients, nephrologists, and dialysis nurses.","workflow_boundary":"Pre-session prescription and within-session adjustment of ultrafiltration and dialysate settings.","failure_mode":"Coordinate-level correction acts on a visible symptom while the outcome-sensitive hemodynamic mode remains amplified.","unit_of_analysis":"Patient–dialysis session.","causal_lever":"Choose the feasible setting change with greatest predicted damping of the hypotension-sensitive mode.","archetype_mapping":"Factor the setting-to-physiology response map with SVD, sweep feasible controls in modal coordinates, and retain only recommendations passing out-of-sample reconstruction checks.","expected_value":"Potential reduction in symptomatic hypotension without sacrificing dialysis adequacy, if individual response modes repeat across sessions.","falsifiable_claim":"Compared with usual adjustment, mode-guided settings will reduce symptomatic hypotension sessions by at least 20% while changing delivered Kt/V by no more than 0.1.","diversity_rationale":"Uses control sensitivity within a repeated extracorporeal treatment session, differing from diagnostic monitoring, network allocation, simulation, and motor retraining.","mechanism_slugs":["singular_value_decomposition","modal_sensitivity_sweep","residual_reconstruction_test"],"search_questions":["Has modal or latent-response control been tested for dialysis prescription adjustment?","How many prior sessions are needed to estimate stable patient-specific response modes?","Which machine settings can independently influence the identified modes within safety limits?"]},{"hypothesis_id":"H3","title":"Spectral targeting across care-transfer networks","problem":"Equal or census-based infection-prevention allocation may overlook facilities that structurally amplify transmission through patient transfers.","affected_stakeholder":"Residents, infection-prevention teams, and regional public-health allocators.","workflow_boundary":"Weekly allocation of limited screening, prophylaxis, vaccination, or outbreak-support capacity across connected facilities.","failure_mode":"Local case counts under-rank a facility embedded in the dominant transfer-network mode.","unit_of_analysis":"Facility–week within a regional transfer network.","causal_lever":"Prioritize limited preventive capacity to facilities with high dominant-mode participation, subject to equity and minimum-service constraints.","archetype_mapping":"Treat patient-transfer connectivity as the operator, estimate its dominant eigenvector, and convert node entries into a constrained intervention ranking.","expected_value":"Potentially fewer cross-facility introductions per unit of scarce preventive capacity when transfer structure predicts spread.","falsifiable_claim":"With the same prevention budget, spectral targeting will reduce secondary facility introductions by at least 15% versus census-based allocation in prospective simulation or a stepped-wedge trial.","diversity_rationale":"Changes regional resource allocation using network position; its stakeholder, boundary, failure mechanism, and facility-week unit differ from patient-level hypotheses.","mechanism_slugs":["network_spectral_centrality_analysis","power_iteration_probe","spectral_gap_monitor"],"search_questions":["Which pathogens and care-transfer settings show transmission aligned with spectral centrality?","Does spectral targeting outperform census, degree, and recent-incidence rules under realistic constraints?","How stable and equitable is the ranking when transfer data are delayed or incomplete?"]},{"hypothesis_id":"H4","title":"Respiratory-mode surrogate for adaptive radiotherapy","problem":"Full deformable dose recalculation may be too slow for routine adaptation to respiratory anatomy.","affected_stakeholder":"Patients receiving thoracic radiotherapy, dosimetrists, and radiation oncologists.","workflow_boundary":"Daily imaging through gating, beam, or replanning decision before treatment delivery.","failure_mode":"A fixed geometric margin misses coupled tumor–organ motion, while a full model cannot return a decision inside the treatment window.","unit_of_analysis":"Patient–treatment fraction.","causal_lever":"Select gating phase or plan adjustment using a fast surrogate that retains dose-relevant deformation modes.","archetype_mapping":"Use SVD-derived deformation modes to construct a reduced-order model, then reject its recommendation when reconstructed anatomy or dose exceeds a prespecified residual budget.","expected_value":"Potentially faster adaptive decisions with bounded dosimetric error, provided respiratory modes remain valid for the fraction.","falsifiable_claim":"The surrogate will cut median adaptive calculation time by at least 40% while keeping organ-at-risk dose error within 2% and target coverage noninferior to full calculation.","diversity_rationale":"Produces a computational artifact for per-fraction planning rather than an alert, allocation ranking, treatment-setting controller, or physical retraining protocol.","mechanism_slugs":["singular_value_decomposition","reduced_order_model","residual_reconstruction_test"],"search_questions":["Which respiratory-motion reduced-order models already support clinical radiotherapy decisions?","What residual threshold preserves target and organ-at-risk dose accuracy?","How often do mode drift or nonrespiratory deformation invalidate the surrogate?"]},{"hypothesis_id":"H5","title":"Empirical gait modes for stroke rehabilitation","problem":"Joint-by-joint rehabilitation may miss coupled gait patterns that constrain recovery after stroke.","affected_stakeholder":"Stroke survivors, physical therapists, and rehabilitation physicians.","workflow_boundary":"Initial movement assessment through selection and progression of gait exercises or robotic assistance.","failure_mode":"Therapy strengthens an obvious weak coordinate while a coupled balance or propulsion mode remains unstable or insensitive to that exercise.","unit_of_analysis":"Patient–rehabilitation course, with session-level mode measurements.","causal_lever":"Apply assistance, resistance, or cueing along the safely measured gait mode with the greatest outcome sensitivity.","archetype_mapping":"Recover patient-specific mode shapes from instrumented walking and safe perturbations, sweep candidate therapy inputs for leverage, and bound interpretation to tested speeds and supports.","expected_value":"Potentially more efficient functional recovery if empirical modes are repeatable and can be manipulated safely.","falsifiable_claim":"After six weeks, mode-targeted therapy will improve six-minute walk distance at least 15% more than impairment-matched conventional therapy without increasing falls or near-falls.","diversity_rationale":"Uses empirical excitation and a course-level motor outcome; unlike the other hypotheses, the operator is unknown and modes are measured directly from physical response.","mechanism_slugs":["mode_shape_testing","modal_sensitivity_sweep","spectral_decomposition_report"],"search_questions":["Have patient-specific gait modes already been used to prescribe post-stroke therapy?","Are measured mode shapes reliable across sessions, speeds, fatigue, and assistive devices?","Can available therapy inputs selectively alter a target mode without harmful cross-coupling?"]}]}