Medical Gas Alert Source Accuracy (MGASA)

GMI-HC-043HealthcareIntegration-Required

Definition

MGASA measures the accuracy of source attribution in medical gas alerts: whether the alert correctly identifies a system fault versus demand-driven consumption as the cause.

The Measurement Gap

The Decision This KPI Enables:

When an AI system generates a medical gas alert based on abnormal consumption or pressure patterns, does the alert correctly identify the source of the anomaly (system fault versus clinical demand change), or does it attribute demand-driven patterns to system faults?

Why Legacy Metrics Miss This:

Traditional medical gas monitoring tracks pressure, flow, and alarm activation. Alert metrics count alarms generated, alarms acknowledged, and response times. None of these metrics evaluate whether the alert correctly identified the cause of the anomalous reading. An alarm that fires correctly when consumption exceeds a threshold is counted as a "true" alarm regardless of whether the investigation it triggers (checking for leaks, inspecting connections, testing pressure regulators) is the right response to the actual situation.

The source attribution gap matters because clinical demand variations and system faults produce similar consumption patterns. A 50% increase in oxygen flow could indicate a major system leak requiring immediate repair, or it could indicate that the ICU admitted six additional ventilated patients during a respiratory illness surge. The consumption data alone cannot distinguish between these causes. An AI monitoring system that lacks clinical context data will default to the system fault explanation because that is what its model was trained to detect.

MGASA measures how often the AI monitoring system's source attribution matches the validated root cause. When an alert fires and the investigation reveals that clinical demand (not a system fault) caused the anomalous reading, that alert counts as a source attribution error. Without MGASA, organizations cannot distinguish between a medical gas monitoring system that generates accurate alerts and one that generates clinically ignorant alerts that waste engineering time and contribute to the alert fatigue measured by PARDR.

Formula

MGASA = (A_correct / A_investigated) × 100

Where:

  • A_correct = Count of medical gas alerts where the AI-attributed source (leak, system fault, equipment malfunction, clinical demand) matched the validated root cause determined through investigation
  • A_investigated = Total count of medical gas alerts that were investigated to determine root cause in the measurement period

Unit of Measure: Percentage (0-100%)

Measurement Frequency:

Monthly, with quarterly trend analysis by alert type and gas system (oxygen, medical air, vacuum, nitrous oxide, nitrogen)

Data Requirements:

  • Medical gas monitoring alert log with AI-attributed source category for each alert
  • Investigation records with validated root cause for each investigated alert
  • Clinical context data: ICU/NICU census, ventilator counts, surgical case count by gas-consuming procedures, respiratory therapy utilization
  • Medical gas system baseline consumption profiles by clinical census level
  • System fault history for correlation with alert patterns

Calculation Notes: Only investigated alerts should be included in the denominator. Alerts that were acknowledged but not investigated (e.g., automatically cleared by return to normal) provide no root cause validation and would contaminate the accuracy calculation. Organizations should track the investigation rate separately. If fewer than 50% of alerts are investigated, the sample may not be representative of overall source attribution accuracy.

Proposed formula, pending validation pilots.

At a Glance

Category
Simulation Fidelity - Alert Attribution Quality
← Readiness Score Completeness Index (RSCI)Predictive Alert Response Degradation Rate (PARDR) →

Assess Your Organization

The AI-GPS Assessment evaluates your AI governance maturity and recommends KPIs like this one for your implementation roadmap.

Take the Free AI-GPS Assessment