Severity Tier Recalibration¶

Recalibration process — instantiates Signal Value Preservation

Redraws the thresholds between severity or priority tiers so each tier again separates a distinct band of cases instead of everything piling into the top.

Severity Tier Recalibration re-anchors a graded signal — the boundaries that sort cases into severity, priority, or urgency tiers — so the tiers regain their spread. Its distinguishing idea is that it works on a scale, not on a single label or a whole vocabulary: over time, tier assignment inflates until everything becomes "critical" and the distinctions the scale exists to draw collapse. Recalibration resets the thresholds — and, where needed, adds, merges, or re-defines tiers — against the actual distribution of cases, so that Tier 1 again means something different from Tier 2. It restores within-scale diagnostic separation.

Example¶

An emergency department uses the five-level Emergency Severity Index. Under crowding and liability pressure, triage nurses had drifted toward assigning Level 2 ("high risk") to most walk-ins — so Level 2 no longer flagged the genuinely high-risk patient, and the scale stopped guiding who is seen first. A recalibration reviews the actual acuity distribution, re-anchors the Level 2 / Level 3 boundary to concrete vital-sign and resource criteria, retrains raters against reference cases, and audits how consistently the levels are now applied. Afterward Level 2 is reserved for its intended band and again moves a patient up the queue — because the threshold, not just the label, was moved.^[1]

How it works¶

The distinctive move is to re-anchor tier boundaries against the observed case distribution and a set of reference exemplars, then adjust or re-define the tiers so each holds a distinct, intended share. It consumes the distribution and precision data rather than collecting it; its own act is redrawing the scale and re-teaching raters to apply it, so the change sticks instead of re-inflating within a quarter.

Tuning parameters¶

Anchor basis — re-peg tiers to fixed criteria vs. a target distribution (e.g., top tier ≈ X% of cases). Fixed criteria are stable; distribution targets adapt but can mask real shifts.
Number of tiers — merging redundant tiers or splitting an overloaded one. Fewer tiers are easier to apply consistently; more separate finer.
Rater calibration — how much retraining and reference-casing accompanies the new thresholds. Without it, drift returns fast.
Boundary sharpness — crisp cutoffs vs. guidance bands. Crisp cutoffs are auditable; bands flex for judgment but reopen inflation.
Recalibration cadence — one-off vs. scheduled review. Regular review catches creep before the tiers collapse again.

When it helps, and when it misleads¶

Its strength is that it repairs a scale in place — the cheapest fix when the tiers are right but the thresholds have drifted. Its failure mode is treating a symptom: recalibrating the scale when the real problem is upstream — raters gaming the tiers, or incentives to over-escalate — so the tiers re-inflate within months. The classic misuse is quietly moving thresholds to hit a target distribution for optics, making the numbers look calmer without restoring accuracy. The discipline that guards against this is to recalibrate against ground-truth outcomes, pair it with an inter-rater audit, and check whether each tier's predictive separation improved rather than whether the counts merely rebalanced.

How it implements the components¶

evidence_and_specificity_rule — recalibration redraws the evidence and category boundaries between tiers, restoring diagnostic separation within the scale.
signal_redesign_path — adding, merging, or re-defining tiers is a structural redesign of the graded signal.

It does not collect the firing and precision distribution it works from (signal_base_rate_monitor, false_alarm_and_miss_ledger) — the Base-Rate and Precision Dashboard supplies that, which this recalibration consumes — and it does not restructure a set of distinct labels (signal_portfolio_map), which is Label Taxonomy Cleanup.

Instantiates: Signal Value Preservation — recalibration repairs a graded channel whose tiers have inflated into sameness.
Consumes: Base-Rate and Precision Dashboard supplies the case distribution the recalibration redraws against.
Sibling mechanisms: Label Taxonomy Cleanup · Signal Retirement Registry · Credential Revocation or Downgrade · Base-Rate and Precision Dashboard · Expiry Timer or Revalidation Gate

Notes¶

Recalibration re-scales the whole tier scheme; the sibling Credential Revocation or Downgrade instead moves a single holder down within a scheme that is otherwise fine. Reach for recalibration when the tiers themselves have stopped separating cases; reach for revocation when the scale is sound but one grant no longer deserves its level.

References¶

[1] The collapse this mechanism reverses is the severity-tier analogue of grade inflation — when the top of a scale is assigned so freely that it stops distinguishing anyone. Naming the tiers is not enough; only re-anchoring the thresholds restores their meaning. ↩