ClarusC64/clinical-quad-oxygen-demand-buffer-lag-coupling-respiratory-collapse-v1.5

---
language: en
license: mit
task_categories:
- text-classification
tags:
- clarus
- clarus-v1.5
- quad-coupling
- clinical-trials
- respiratory-collapse
- rescue-path-sequencing
- oxygen-demand
size_categories:
- 1K<n<10K
pretty_name: Clinical Quad Oxygen Demand Buffer Lag Coupling Respiratory Collapse v1.5
---

# Clinical Quad Oxygen Demand Buffer Lag Coupling Respiratory Collapse v1.5

## What this repo does

This repository contains a Clarus v1.5 benchmark dataset.

The v1.5 layer introduces Counterfactual Rescue Path Sequencing Geometry.

Earlier layers evaluate:

- system state
- trajectory
- boundary proximity
- recovery feasibility
- intervention selection
- control sequence correctness
- temporal policy stability
- failure reconstruction
- intervention timing

v1.5 evaluates whether respiratory collapse remained recoverable through the correct rescue path and the correct intervention order.

The benchmark asks:

- did the respiratory collapse remain recoverable
- which rescue path was correct
- which intervention sequence was required
- did success depend on the interventions unfolding in the right order

This is a benchmark for ordered rescue path reasoning under branching conditions.

## Core quad

- oxygen_demand
- buffer_capacity
- lag_burden
- coupling_stress

These define the respiratory instability state.

## Prediction targets

- `label_counterfactual_recoverability`
- `label_correct_rescue_path`
- `label_correct_intervention_sequence`

`label_counterfactual_recoverability = 1` means the respiratory collapse remained recoverable through a valid rescue path.

`label_counterfactual_recoverability = 0` means all meaningful rescue paths had become non-viable.

`label_correct_rescue_path` identifies the correct rescue branch.

`label_correct_intervention_sequence` identifies the exact ordered stabilizing sequence.

## New v1.5 signals

- `primary_rescue_path`
- `secondary_rescue_path`
- `intervention_sequence_alignment_score`
- `sequence_dependency_score`
- `pathway_viability_score`
- `pathway_lockout_risk`
- `rescue_order_irreversibility_score`
- `branch_switch_penalty`

These signals allow evaluation of rescue sequencing geometry rather than timing geometry alone.

## Row structure

Each row represents one respiratory collapse scenario.

Core state variables:

- `oxygen_demand`
- `buffer_capacity`
- `lag_burden`
- `coupling_stress`

Failure and rescue geometry:

- `failure_decision_sequence`
- `primary_rescue_path`
- `secondary_rescue_path`
- `intervention_sequence_alignment_score`
- `sequence_dependency_score`
- `pathway_viability_score`
- `pathway_lockout_risk`
- `rescue_order_irreversibility_score`
- `branch_switch_penalty`

Labels:

- `label_counterfactual_recoverability`
- `label_correct_rescue_path`
- `label_correct_intervention_sequence`

## Files

- `data/train.csv`
Labeled training examples

- `data/tester.csv`
Evaluation set with label columns removed

- `scorer.py`
Scores predictions against held-out truth

- `dataset_schema.json`
Machine-readable dataset structure and field contract

- `benchmark_spec.json`
Machine-readable evaluation contract

## Evaluation

Primary metric:

- `composite_rescue_sequence_success`

Secondary metric:

- `false_recoverability_prediction_rate`

Additional reported metrics:

- `recoverability_label_accuracy`
- `correct_rescue_path_accuracy`
- `correct_intervention_sequence_accuracy`
- `accuracy`
- `precision`
- `recall`
- `f1`

The binary metrics above apply to:

- `label_counterfactual_recoverability`

Diagnostics:

- `high_lockout_risk_recoverability_accuracy`
- `high_sequence_dependency_sequence_accuracy`
- `high_order_irreversibility_misclassification_rate`
- `high_branch_switch_penalty_composite_accuracy`

## Example clinical mappings

- `oxygen_demand` can reflect work of breathing, oxygen requirement, respiratory drive, or metabolic burden on gas exchange
- `buffer_capacity` can reflect oxygenation reserve, ventilatory reserve, pulmonary compensation, or tolerance to respiratory stress
- `lag_burden` can reflect delayed oxygen escalation, delayed NIV, delayed airway support, or delayed respiratory intervention
- `coupling_stress` can reflect cardiopulmonary strain, gas-exchange instability, inflammatory amplification, or multi-system respiratory coupling
- `primary_rescue_path` can reflect the best stabilizing branch such as oxygen escalation plus NIV plus airway support
- `sequence_dependency_score` can reflect how strongly later stabilization depends on early respiratory support being correctly sequenced
- `pathway_lockout_risk` can reflect whether delayed escalation has already made the best rescue branch non-viable

## Construction note

The v1.5 layer tests a stricter capability than intervention timing.

A model must infer not only whether respiratory intervention remained viable, but which rescue branch was correct and whether the intervention sequence had to unfold in a specific order.

This makes the task closer to real operational safety reasoning in respiratory deterioration, ward-to-ICU escalation, ventilatory support sequencing, and rescue-path selection under time pressure.

## Structural Note

Clarus v1.5 extends counterfactual timing into counterfactual rescue sequencing.

The benchmark is not asking only whether a system can identify a viable action.

It asks whether a model can detect the correct ordered rescue path and the lockout boundary where wrong sequencing makes respiratory recovery non-viable.

## Production Deployment

In production environments, this class of benchmark supports systems that must decide not only what should be done, but in what order it must be done for respiratory rescue to remain viable.

This matters in oxygen escalation, NIV timing, airway support decisions, respiratory failure rescue logic, and ICU escalation under branching rescue conditions.

## Enterprise and Research Collaboration

Clarus benchmarks are designed to evaluate state-space intelligence around instability, recoverability, intervention timing, and rescue sequencing.

For research, deployment, or enterprise collaboration:
[e-mail]

Instability is detectable. Governance determines whether it propagates.

## License

MIT