Risk Score Action Rate (RSAR)

GMI-HC-014HealthcareIntegration-Required

Definition

Percentage of patients with high-risk scores (above a defined threshold) who receive a documented intervention attributable to the risk score, distinguishing between risk models that drive action and risk models that decorate the chart.

The Measurement Gap

The Decision That Legacy Metrics Cannot Support

Is the readmission risk model actually changing care plans, or are clinicians seeing the score, acknowledging it, and doing nothing different?

What Goes Unmeasured Today

Healthcare organizations have deployed risk prediction models across multiple clinical domains: readmission risk, sepsis risk, fall risk, pressure injury risk, and deterioration risk. These models generate scores that appear in the EHR, on dashboards, and sometimes in clinical alerts. Healthcare organizations measure whether the models are accurate (does a high-risk score predict the event?) and whether clinicians see the scores (are the scores displayed in the workflow?). What they rarely measure is whether the risk scores actually change clinical behavior.

The distinction between a risk model that drives action and a risk model that decorates the chart is the difference between a useful clinical tool and an expensive IT project. A readmission risk model that identifies 200 high-risk patients per month but triggers care management referrals for only 30 of them is 85% wallpaper. The 85% of patients who were flagged but received no additional intervention gained nothing from the model: the organization invested in risk prediction but did not invest in the action infrastructure to respond to it.

Legacy quality metrics (readmission rate, fall rate, sepsis mortality) measure outcomes without connecting them to risk model utilization. If the readmission rate remains at 15% despite deploying a risk model, was the model inaccurate? Were the interventions ineffective? Or did nobody act on the scores? Without measuring the action rate, the organization cannot diagnose which link in the chain is broken.

The Invisible Failure Mode

Risk models that drive no action are invisible cost centers. They consume IT resources to deploy and maintain, appear on dashboards that leadership monitors, and generate scores that clinicians acknowledge. The absence of resulting clinical action is invisible because no metric tracks it. The organization may believe the risk model is working because it exists in the workflow, never asking whether it changes anything.

Formula

Primary Formula

RSAR = Patients_high_risk_with_intervention / Patients_high_risk_total

Variable Definitions

Variable Definition Unit
Patients_high_risk_total Patients whose risk score exceeded the actionable threshold during the measurement period Count
Patients_high_risk_with_intervention Subset who received a documented intervention linked to the risk score Count
Intervention linkage Intervention documented within a defined time window of the risk score AND intervention type matches the risk domain Temporal + domain match
Risk domains Readmission (care management referral), fall (prevention protocol), sepsis (workup initiation), pressure injury (prevention bundle) Category

Data Requirements

Data Element Source System Classification
Risk scores by model and patient EHR risk model output NATIVE
Risk score threshold configurations EHR clinical decision support rules NATIVE
Documented interventions EHR: orders, referrals, care plans, flowsheet entries NATIVE
Intervention-to-risk-score linkage Derived: temporal proximity + intervention type matching risk domain DERIVED

Edge Cases and Exclusions

Proposed formula, pending validation pilots.

At a Glance

Category
AI Governance - Risk Model Governance
← Risk Model Discrimination Currency (RMDC)Assignment Competency-Acuity Match Index (ACAMI) →

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